This paper examines the explanatory gap account. The key notions for its proper understanding are analysed. In particular, the analysis is concerned with the role of “thick” and “thin” modes of presentation and “thick” and “thin” concepts which are relevant for the notions of “thick” and “thin” conceivability, and to that effect relevant for the gappy and non-gappy identities. The last section of the paper discusses the issue of the intelligibility of explanations. One of the conclusions is that the (...) explanatory gap account only succeeds in establishing the epistemic gap. The claim that psychophysical identity is not intelligibly explicable, and thus opens the explanatory gap, would require an indepen- dent argument which would prove that intelligible explanations stem only from conceptual analysis. This, I argue, is not the case. (shrink)

Many philosophers of science believe that empirical psychology can contribute little to the philosophical investigation of explanations. They take this to be shown by the fact that certain explanations fail to elicit any relevant psychological events (e.g., familiarity, insight, intelligibility, etc.). We report results from a study suggesting that, at least among those with extensive science training, a capacity to render an event intelligible is considered a requirement for explanation. We also investigate for whom explanations must be capable (...) of rendering events intelligible and whether or not accuracy is also viewed as a requirement. (shrink)

Here is my thesis (and the outline of this paper). Increasingly secret, complex and inscrutable computational systems are being used to intensify existing power relations, and to create new ones (Section II). To be all-things-considered morally permissible, new, or newly intense, power relations must in general meet standards of procedural legitimacy and proper authority (Section III). Legitimacy and authority constitutively depend, in turn, on a publicity requirement: reasonably competent members of the political community in which power is being exercised must (...) be able to determine that power is being exercised legitimately and with proper authority (Section IV). The publicity requirement can be satisfied only if the powerful can explain their decision-making—including the computational tools that they use to support it—to members of their political community. Section V applies these ideas to opaque computational systems. Section VI addresses objections; Section VII concludes. (shrink)

Artificial intelligence (AI) and robotics are digital technologies that will have significant impact on the development of humanity in the near future. They have raised fundamental questions about what we should do with these systems, what the systems themselves should do, what risks they involve, and how we can control these. - After the Introduction to the field (§1), the main themes (§2) of this article are: Ethical issues that arise with AI systems as objects, i.e., tools made and used (...) by humans. This includes issues of privacy (§2.1) and manipulation (§2.2), opacity (§2.3) and bias (§2.4), human-robot interaction (§2.5), employment (§2.6), and the effects of autonomy (§2.7). Then AI systems as subjects, i.e., ethics for the AI systems themselves in machine ethics (§2.8) and artificial moral agency (§2.9). Finally, the problem of a possible future AI superintelligence leading to a “singularity” (§2.10). We close with a remark on the vision of AI (§3). - For each section within these themes, we provide a general explanation of the ethical issues, outline existing positions and arguments, then analyse how these play out with current technologies and finally, what policy consequences may be drawn. (shrink)

Since approval of the EU General Data Protection Regulation (GDPR) in 2016, it has been widely and repeatedly claimed that the GDPR will legally mandate a ‘right to explanation’ of all decisions made by automated or artificially intelligent algorithmic systems. This right to explanation is viewed as an ideal mechanism to enhance the accountability and transparency of automated decision-making. However, there are several reasons to doubt both the legal existence and the feasibility of such a right. In contrast (...) to the right to explanation of specific automated decisions claimed elsewhere, the GDPR only mandates that data subjects receive meaningful, but properly limited, information (Articles 13-15) about the logic involved, as well as the significance and the envisaged consequences of automated decision-making systems, what we term a ‘right to be informed’. Further, the ambiguity and limited scope of the ‘right not to be subject to automated decision-making’ contained in Article 22 (from which the alleged ‘right to explanation’ stems) raises questions over the protection actually afforded to data subjects. These problems show that the GDPR lacks precise language as well as explicit and well-defined rights and safeguards against automated decision-making, and therefore runs the risk of being toothless. We propose a number of legislative and policy steps that, if taken, may improve the transparency and accountability of automated decision-making when the GDPR comes into force in 2018. (shrink)

Artificial intelligence (AI) is a digital technology that will be of major importance for the development of humanity in the near future. AI has raised fundamental questions about what we should do with such systems, what the systems themselves should do, what risks they involve and how we can control these. - After the background to the field (1), this article introduces the main debates (2), first on ethical issues that arise with AI systems as objects, i.e. tools made and (...) used by humans; here, the main sections are privacy (2.1), manipulation (2.2), opacity (2.3), bias (2.4), autonomy & responsibility (2.6) and the singularity (2.7). Then we look at AI systems as subjects, i.e. when ethics is for the AI systems themselves in machine ethics (2.8.) and artificial moral agency (2.9). Finally we look at future developments and the concept of AI (3). For each section within these themes, we provide a general explanation of the ethical issues, we outline existing positions and arguments, then we analyse how this plays out with current technologies and finally what policy conse-quences may be drawn. (shrink)

Explanation is a foundational goal in the exact sciences. Besides the contemporary considerations on ‘description’, ‘classification’, and ‘prediction’, we often see these terms in thriving applications of artificial intelligence (AI) in chemistry hypothesis generation. Going beyond describing ‘things in the world’, these applications can make accurate numerical property calculations from theoretical or topological descriptors. This association makes an interesting case for a logic of discovery in chemistry: are these induction-led ventures showing a shift in how chemists can problematize research (...) questions? In this article, I present a fresh perspective on the current context of discovery in chemistry. I argue how data-driven statistical predictions in chemistry can be explained as a quasi-logical process for generating chemical theories, beyond the classic examples of organic and theoretical chemistry. Through my position on formal models of scientific explanation, I demonstrate how the dawn of AI can provide novel insights into the explanatory power of scientific endeavors. (shrink)

In this article, we argue that there is an essential difference between social intelligence and creative intelligence, and that they have their foundation in human sexuality. For sex differences, we refer to the vast psychological, neurological, and cognitive science research where problem-solving, verbal skills, logical reasoning, and other topics are dealt with. Intelligence tests suggest that, on average, neither sex has more general intelligence than the other. Though people are equals in general intelligence, they are different in special forms of (...) intelligence such as social intelligence and creative intelligence, the former dominant in women, the latter dominant in men. The dominance of creative intelligence in men needs to be explained. The focus of our research is on the strictly anthropological aspects, and consequently our explanation for this fact is based on the male-female polarity in the mating systems. Sexual dimorphism does not only regard bodily differences but implies different forms of sex life. Sex researchers distinguish between two levels of sexual intercourse: procreative sex and recreational sex, and to these we would add “creative sex.” On all three levels, there is a behavioral difference between men and women, including the subjective experience. These differences are as well attributed to culture as genetically founded in nature. Sexual reproduction is only possible if females cooperate. Their biological inheritance makes females play a decisive role in mate choice. Recreational sex for the purpose of pleasure rather than reproduction results from female extended sexual activity. Creative sex, on the contrary, is a specifically male performance of sexuality. We identify creative sex with eroticism. Eroticism evolved through the transformation of the sexual drive into a mental state of expectation and fantasizing. Hence, sex differences (that nowadays are covered up by cultural egalitarianism) continue to be the evolutionary origin of the difference between social and creative intelligence. (shrink)

Recent work on interpretability in machine learning and AI has focused on the building of simplified models that approximate the true criteria used to make decisions. These models are a useful pedagogical device for teaching trained professionals how to predict what decisions will be made by the complex system, and most importantly how the system might break. However, when considering any such model it’s important to remember Box’s maxim that "All models are wrong but some are useful." We focus on (...) the distinction between these models and explanations in philosophy and sociology. These models can be understood as a "do it yourself kit" for explanations, allowing a practitioner to directly answer "what if questions" or generate contrastive explanations without external assistance. Although a valuable ability, giving these models as explanations appears more difficult than necessary, and other forms of explanation may not have the same trade-offs. We contrast the different schools of thought on what makes an explanation, and suggest that machine learning might benefit from viewing the problem more broadly. (shrink)

Review of Henk de Regt's "Understanding Scientific Understanding".

Against ideas of dualism, logocentrism, anthropocentrism, animism, panpsychism, biocentrism, neurocentrism, foundationalism, computationalism, especially substantialism, reductionism and even physicalism, the author argues that life may be the only non-reductive concept, even the only ontological concept, with which we may explain our consciousness, intelligence and language. Life, as defined in this article, explains but not only human brains, and even not only biological organisms. Still, the mind, also as defined in this article, is the only one it explains. No mind may exist (...) if not be a life or lives, and no life may exist if not be a mind or a part of it. If it is the mind that needs to be explained, it must finally and fundamentally be explained as a life or lives. If the question is about the origin of the mind, a life or lives must be the ultimate answer. In other words, life is the only attribute of mind, and mind also the only attribute of life, and therefore, consciousness, intelligence and language must be the properties of all the living systems, including non-biological living systems. A model of mind is hypothesized based on the analysis of two kinds of lives and their relationship with matter and energy. It may be deduced from this model: 1. All the memories of a living brain are its intelligence. 2. Both consciousness and nonconscious are only meaningless languages used in the communication among lives in an awakened brain. 3. Life is the only meaning of all those memories and languages. Ontologically, life may even be the only meaning of all matter and energy. -/- (A version of this article was published by the Journal of NeuroPhilosophy on 26.10.2022). (shrink)

A story does more than recount events; it recounts events in a way that renders them intelligible, thus conveying not just information but also understanding. We might therefore be tempted to describe narrative as a genre of explanation. When the police invite a suspect to “tell his story,” they are asking him to explain the blood on his shirt or his absence from home on the night of the murder; and whether he is judged to have a “good story” (...) will depend on its adequacy as an explanation. Can we account for the explanatory force of narrative with the models of explanation available in the philosophy of science? Or does narrative convey a different kind of understanding, which requires a different model and perhaps even a term other than ‘explanation’? (shrink)

The burden of research ethics review falls not just on researchers, but on those who serve on research ethics committees (RECs). With the advent of automated text analysis and generative artificial intelligence, it has recently become possible to teach models to support human judgment, for example by highlighting relevant parts of a text and suggesting actionable precedents and explanations. It is time to consider how such tools might be used to support ethics review and oversight. This commentary argues that with (...) a suitable strategy of engagement, artificial intelligence can be used in a variety of ways that can genuinely support RECs and manage their workload and improve the quality of review. It would be wiser to take an active role in the development of AI tools for ethics review, rather than to adopt ad hoc tools after the fact. (shrink)

Endowing artificial systems with explanatory capacities about the reasons guiding their decisions, represents a crucial challenge and research objective in the current fields of Artificial Intelligence (AI) and Computational Cognitive Science [Langley et al., 2017]. Current mainstream AI systems, in fact, despite the enormous progresses reached in specific tasks, mostly fail to provide a transparent account of the reasons determining their behavior (both in cases of a successful or unsuccessful output). This is due to the fact that the classical problem (...) of opacity in artificial neural networks (ANNs) explodes with the adoption of current Deep Learning techniques [LeCun, Bengio, Hinton, 2015]. In this paper we argue that the explanatory deficit of such techniques represents an important problem, that limits their adoption in the cognitive modelling and computational cognitive science arena. In particular we will show how the current attempts of providing explanations of the deep nets behaviour (see e.g. [Ritter et al. 2017] are not satisfactory. As a possibile way out to this problem, we present two different research strategies. The first strategy aims at dealing with the opacity problem by providing a more abstract interpretation of neural mechanisms and representations. This approach is adopted, for example, by the biologically inspired SPAUN architecture [Eliasmith et al., 2012] and by other proposals suggesting, for example, the interpretation of neural networks in terms of the Conceptual Spaces framework [Gärdenfors 2000, Lieto, Chella and Frixione, 2017]. All such proposals presuppose that the neural level of representation can be considered somehow irrelevant for attacking the problem of explanation [Lieto, Lebiere and Oltramari, 2017]. In our opinion, pursuing this research direction can still preserve the use of deep learning techniques in artificial cognitive models provided that novel and additional results in terms of “transparency” are obtained. The second strategy is somehow at odds with respect to the previous one and tries to address the explanatory issue by avoiding to directly solve the “opacity” problem. In this case, the idea is that one of resorting to pre-compiled plausible explanatory models of the word used in combination with deep-nets (see e.g. [Augello et al. 2017]). We argue that this research agenda, even if does not directly fits the explanatory needs of Computational Cognitive Science, can still be useful to provide results in the area of applied AI aiming at shedding light on the models of interaction between low level and high level tasks (e.g. between perceptual categorization and explanantion) in artificial systems. (shrink)

Abstract. Intelligent Tutoring System Builder (ITSB) is an authoring tool designed and developed to aid teachers in constructing intelligent tutoring systems in a multidisciplinary fields. The teacher is needed to create a set of pedagogical fundamentals, which, in line, are inured to automatically build up a broad tutor framework and construct an intelligent tutoring system. In this paper an explanation of the theory and the architecture of the tool is outlined. A presentation of several system components, the requirements of (...) the different components, integration of these components in ITSB tool are shown. Furthermore, implanting of requirements, cognitive principle, and common design fundamentals in the tool to ease the use of teachers. A variety of design matters, an example of building an intelligent tutoring system for teaching Java language using ITSB tool and an evaluation are presented. (shrink)

In a recent article, Erasmus, Brunet, and Fisher (2021) argue that Artificial Neural Networks (ANNs) are explainable. They survey four influential accounts of explanation: the Deductive-Nomological model, the Inductive-Statistical model, the Causal-Mechanical model, and the New-Mechanist model. They argue that, on each of these accounts, the features that make something an explanation is invariant with regard to the complexity of the explanans and the explanandum. Therefore, they conclude, the complexity of ANNs (and other Machine Learning models) does not (...) make them less explainable. In this reply, it is argued that Erasmus et al. left out one influential account of explanation from their discussion: the Unificationist model. It is argued that, on the Unificationist model, the features that makes something an explanation is sensitive to complexity. Therefore, on the Unificationist model, ANNs (and other Machine Learning models) are not explainable. It is emphasized that Erasmus et al.’s general strategy is correct. The literature on explainable Artificial Intelligence can benefit by drawing from philosophical accounts of explanation. However, philosophical accounts of explanation do not settle the problem of whether ANNs are explainable because they do not unanimously declare that explanation is invariant with regard to complexity. (shrink)

The author argues that consciousness may never be defined or explained with entities or their properties. As an alternative, the author proposes a system with matter, energy, and lives as its components, and with all its components defined as changes. Based on the relationships among these three components, a triple definition or explanation of consciousness is reached: • Ontologically, consciousness is universal, since it is the distinction between matter and energy. • Epistemologically, consciousness is unique, since it is the (...) energy formalized, qualified, or diversified by the matter. • Semantically, consciousness is a meaningless language with lives as its only meaning. Many theories of consciousness have neglected such a universality or uniqueness and even ignored lives as the only cause or effect of consciousness. -/- . (shrink)

This paper examines a parallel between Aristotle’s account of phantasia and contemporary psychological models of working memory, a capacity that enables the temporary maintenance and manipulation of information used in many behaviors. These two capacities, though developed within two distinct scientific paradigms, share a common strategy of psychological explanation, Aristotelian Faculty Psychology. This strategy individuates psychological components by their target-domains and functional roles. Working memory and phantasia result from an attempt to individuate the psychological components responsible for flexible thought (...) and are thus implicated in most of our robust cognitive processes, from reading comprehension to problem solving. We then present two novel objections which suggest that these capacities cannot explain our ability to engage in flexible thought. To escape the resultant impasse, we survey alternatives and argue that most promising strategies depend on identifying the behaviors attributed to intelligent thought and action. (shrink)

Explainability and comprehensibility of AI are important requirements for intelligent systems deployed in real-world domains. Users want and frequently need to understand how decisions impacting them are made. Similarly it is important to understand how an intelligent system functions for safety and security reasons. In this paper, we describe two complementary impossibility results (Unexplainability and Incomprehensibility), essentially showing that advanced AIs would not be able to accurately explain some of their decisions and for the decisions they could explain people would (...) not understand some of those explanations. (shrink)

In his book “Minds and Computers: An Introduction to the Philosophy of Artificial Intelligence”, Matt Carter presents a comprehensive exploration of the philosophical questions surrounding artificial intelligence (AI). Carter argues that the development of AI is not merely a technological challenge but fundamentally a philosophical one. He delves into key issues like the nature of mental states, the limits of introspection, the implications of memory decay, and the functionalist framework that allows for the possibility of AI. Carter contrasts functionalism with (...) reductive materialism, highlighting how the former accommodates the concept of artificial minds. He also emphasizes the significance of computationalism, which combines functionalism with computational theory to provide a robust explanation for mental processes. The book further discusses formal systems, the role of register machines in computation, and the inherent challenges in AI research, particularly in developing natural language processing systems. Carter’s work underscores the limitations of current computational models of cognition while remaining hopeful that advances in neuroscience may lead to stronger AI systems in the future. (shrink)

The ongoing explosion of interest in artificial intelligence is fueled in part by recently developed techniques in machine learning. Those techniques allow automated systems to process huge amounts of data, utilizing mathematical methods that depart from traditional statistical approaches, and resulting in impressive advancements in our ability to make predictions and uncover correlations across a host of interesting domains. But as is now widely discussed, the way that those systems arrive at their outputs is often opaque, even to the experts (...) who design and deploy them. Is it morally problematic to make use of opaque automated methods when making high-stakes decisions, like whether to issue a loan to an applicant, or whether to approve a parole request? Many scholars answer in the affirmative. However, there is no widely accepted explanation for why transparent systems are morally preferable to opaque systems. We argue that the use of automated decision-making systems sometimes violates duties of consideration that are owed by decision-makers to decision-subjects, duties that are both epistemic and practical in character. Violations of that kind generate a weighty consideration against the use of opaque decision systems. In the course of defending our approach, we show that it is able to address three major challenges sometimes leveled against attempts to defend the moral import of transparency in automated decision-making. (shrink)

When this article was first planned, writing was going to be exclusively about two things - the origin of life and human evolution. But it turned out to be out of the question for the author to restrict himself to these biological and anthropological topics. A proper understanding of them required answering questions like “What is the nature of the universe – the home of life – and how did it originate?”, “How can time travel be removed from fantasy and (...) science fiction, to be made scientific and practical?”, and “How can the proposed young age of genus Homo be made to actually be reasonable – when simply stating it would be solid ground for instant rejection and dismissal?” The result is that the article also talks about subjects like Artificial Intelligence, General Relativity, and cosmology. -/- From where did life originate? God? Evolution? Panspermia? If the tendency of humans and scientists to regard undiscovered science as pseudoscience can be overcome, Einstein gave another alternative to consider when he introduced General Relativity. Time isn’t linear – progressing in a straight line from past to present to future. That assumption ignores Relativity which states that space AND TIME are curved. Where did life and the genetic code come from? Can the answer build AI? -/- The first question can be answered by the section of this article titled SETI, Evolution, and Time which says life (possibly multicellular and intelligent) and the genetic code came from humans acquiring knowledge of these things over the centuries, then applying that knowledge – via terraforming, accumulation of raw materials like amino acids and nucleic acids, genetic engineering - to a time in the past when life didn’t exist. From that origin, life evolved through innumerable mutations and adaptations, with humans once again acquiring knowledge of it in cyclic (nonlinear) time. -/- The second question is answered by saying artificial intelligence (AI) as the product of life is only half of the equation. The other half refers to Relativity’s curved space-time and violation of the notion that time always travels from past to future. We have always lived in an artificially intelligent, non-probabilistic universe where everything in time and space is connected into one thing by quantum entanglement – making the brain and genes products of binary-digit activity or artificial intelligence (life is not merely dependent on biology’s “lock and key” mechanisms but also possesses AI). -/- The earliest documented representative of the genus Homo is Homo habilis, which evolved around 2.8 million years ago. Scientists used to believe there was a straight line from H. habilis to us, Homo sapiens. This article will use the “advanced” waves loved by Physics Nobel laureate Richard Feynman, view the history of science through the lens of Conic Sections applied to Relativity’s curved space-time, and incorporate the necessity of so-called imaginary time * – popularized by Prof. Stephen Hawking. While the evolutionary proposals are more in agreement with this early straight line than with modern theories, Albert Einstein’s General Relativity is used to transform the straight line into a curved line, ultimately concluding that Homo habilis (H. habilis) originated only (and unbelievably, as far as today’s science and technology is concerned) ~250,000 years ago. Other branches and dead ends of Homo – e.g. Neanderthals – are the result of mutations and adaptations, with the resultant modifications to anatomy and physiology. The surprisingly young age of H. habilis allows nearly 200,000 years for habilis, or one of its descendants, to reach Australia … if this country’s indigenous Aboriginal population did, as claimed, reach this “island continent” 60,000 years ago. -/- * The ultraviolet catastrophe, also called the Rayleigh–Jeans catastrophe, is a failure of classical physics to predict observed phenomena: it can be shown that a blackbody - a hypothetical perfect absorber and radiator of energy - would release an infinite amount of energy, contradicting the principles of conservation of energy and indicating that a new model for the behaviour of blackbodies was needed. At the start of the 20th century, physicist Max Planck derived the correct solution by making some strange (for the time) assumptions. In particular, Planck assumed that electromagnetic radiation can only be emitted or absorbed in discrete packets, called quanta. Albert Einstein postulated that Planck's quanta were real physical particles (what we now call photons), not just a mathematical fiction. From there, Einstein developed his explanation of the photoelectric effect (when quanta or photons of light shine on certain metals, electrons are released and can form an electric current). So it appears entirely possible that another supposed mathematical trickery (imaginary time and the y-axis of Wick rotation) will find practical application in the future. -/- The article includes mathematical references to cosmology (spoiler alert – you’ll read about things like Vector-Tensor-Scalar Geometry, topology, the “eternal present”, Einstein’s Unified Field, the inverse-square law, and there being no Big Bang and no multiverse - but there will also be no equations). -/- The other subheadings in this essay are – -/- NONLINEAR TIME AND ELECTRICAL ENGINEERING (about a 2009 electrical engineering experiment at America’s Yale University, and cosmic wormholes) -/- BITS AND TOPOLOGY (base-2 maths aka Binary digiTS, Mobius strips, and figure-8 Klein bottles) -/- WICK ROTATION, CAUSALITY, AND UNITING TIME (do past, present and future co-exist in an “eternal present”?) -/- DIGITAL BRAIN, DIGITAL UNIVERSE (if the brain and the universe are ultimately composed of binary digits, we'll someday be able to do the same things with the brain and universe that we now do with computers) -/- PROPOSAL: HUMAN AND ANIMAL INSTINCTS ARE THE RESULT OF THE UNIVERSE BEING UNIFIED BY BINARY DIGITS (AND TOPOLOGY) (If everything in the universe is ultimately composed of electronic BITS, then the universe must possess Artificial Intelligence - some prefer the term Cosmic Consciousness) -/- INFORMATION THEORY CONQUERS A RED GIANT (preserving Earth by keeping the Sun near today’s level of activity forever). (shrink)

The generation of novel refined scientific conceptions of intelligence, creativity and consciousness is of paramount importance at a time where many scientists deem the technological singularity and the achievement of self-improving superintelligent algorithms to be immanent while numerous other scientists characterize present-day algorithms as the mere implementation of superficial mimicry incapable of yielding outcomes such as superintelligence. The precarious epistemic state of affairs reflected in this discrepancy became increasingly palpable in the unfolding deepfake era even though informed safety- and security-relevant (...) actions need requisite variety in the near-term. Divergently, historically speaking, from the perspective of cosmology, science is described to already have successfully navigated from the geocentric model to the acentric model of the universe representing the currently best known explanation for the structure of the expanding cosmos. In analogy, this autodidactic paper expounds the scientific need for a broader contextualization of the current epistemic situation and explicates why acentric civilization-level epistemic relativity and invariance considerations can enable a more rigorous scientific evaluation of algorithmic superintelligence achievement claims – constraining the latter scientifically. (shrink)

Necessity and sufficiency are the building blocks of all successful explanations. Yet despite their importance, these notions have been conceptually underdeveloped and inconsistently applied in explainable artificial intelligence (XAI), a fast-growing research area that is so far lacking in firm theoretical foundations. Building on work in logic, probability, and causality, we establish the central role of necessity and sufficiency in XAI, unifying seemingly disparate methods in a single formal framework. We provide a sound and complete algorithm for computing explanatory factors (...) with respect to a given context, and demonstrate its flexibility and competitive performance against state of the art alternatives on various tasks. (shrink)

Is there a field of social intelligence? Many various disciplines ap-proach the subject and it may only seem natural to suppose that different fields of study aim at explaining different phenomena; in other words, there is no spe-cial field of study of social intelligence. In this paper, I argue for an opposite claim. Namely, there is a way to integrate research on social intelligence, as long as one accepts the mechanistic account to explanation. Mechanistic inte-gration of different explanations, however, (...) comes at a cost: mechanism requires explanatory models to be fairly complete and realistic, and this does not seem to be the case for many models concerning social intelligence, especially models of economical behavior. Such models need either be made more realistic, or they would not count as contributing to the same field. I stress that the focus on integration does not lead to ruthless reductionism; on the contrary, mechanistic explanations are best understood as explanatorily pluralistic. (shrink)

The opacity of some recent Machine Learning (ML) techniques have raised fundamental questions on their explainability, and created a whole domain dedicated to Explainable Artificial Intelligence (XAI). However, most of the literature has been dedicated to explainability as a scientific problem dealt with typical methods of computer science, from statistics to UX. In this paper, we focus on explainability as a pedagogical problem emerging from the interaction between lay users and complex technological systems. We defend an empirical methodology based on (...) field work, which should go beyond the in-vitro analysis of UX to examine in-vivo problems emerging in the field. Our methodology is also comparative, as it chooses to steer away from the almost exclusive focus on ML to compare its challenges with those faced by more vintage algorithms. Finally, it is also philosophical, as we defend the relevance of the philosophical literature to define the epistemic desiderata of a good explanation. This study was conducted in collaboration with Etalab, a Task Force of the French Prime Minister in charge of Open Data & Open Government Policies, dealing in particular with the enforcement of the right to an explanation. In order to illustrate and refine our methodology before going up to scale, we conduct a preliminary work of case studies on the main different types of algorithms used by the French administration: computation, matching algorithms and ML. We study the merits and drawbacks of a recent approach to explanation, which we baptize input-output black box reasoning or BBR for short. We begin by presenting a conceptual framework including the distinctions necessary to a study of pedagogical explainability. We proceed to algorithmic case studies, and draw model-specific and model-agnostic lessons and conjectures. (shrink)

Modern life sciences research is increasingly relying on artificial intelligence (AI) approaches to model biological systems, primarily centered around the use of machine learning (ML) models. Although ML is undeniably useful for identifying patterns in large, complex data sets, its widespread application in biological sciences represents a significant deviation from traditional methods of scientific inquiry. As such, the interplay between these models and scientific understanding in biology is a topic with important implications for the future of scientific research, yet it (...) is a subject that has received little attention. Here, we draw from an epistemological toolkit to contextualize recent applications of ML in biological sciences under modern philosophical theories of understanding, identifying general principles that can guide the design and application of ML systems to model biological phenomena and advance scientific knowledge. We propose that conceptions of scientific understanding as information compression, qualitative intelligibility, and dependency relation modelling provide a useful framework for interpreting ML-mediated understanding of biological systems. Through a detailed analysis of two key application areas of ML in modern biological research – protein structure prediction and single cell RNA-sequencing – we explore how these features have thus far enabled ML systems to advance scientific understanding of their target phenomena, how they may guide the development of future ML models, and the key obstacles that remain in preventing ML from achieving its potential as a tool for biological discovery. Consideration of the epistemological features of ML applications in biology will improve the prospects of these methods to solve important problems and advance scientific understanding of living systems. (shrink)

This paper serves as an intuition building mechanism for understanding the basics of AI, misalignment, and the reasons for why strong AI is being pursued. The approach is to engage with both pro and anti AI development arguments to gain a deeper understanding of both views, and hopefully of the issue as a whole. We investigate the basics of misalignment, common misconceptions, and the arguments for why we would want to pursue strong AI anyway. The paper delves into various aspects (...) of the problem, including existential risk, deception, hedonic adaptation, and the potential for the complete extinction of humanity. By integrating multiple elements of philosophy, this paper aims to provide a holistic understanding of the alignment problem and its significance for the future of humanity. (shrink)

While a great deal of abuse has been directed at intelligent design theory (ID), its starting point is a fact about biological organisms that cries out for explanation, namely "specified complexity" (SC). Advocates of ID deploy three kind of argument from specified complexity to the existence of a designer: an eliminative argument, an inductive argument, and an inference to the best explanation. Only the first of these merits the abuse directed at it; the other two arguments are worthy (...) of respect. If they fail, it is only because we have a better explanation of SC, namely Darwin's theory of evolution by natural selection. (shrink)

In the case of AI, automated systems are making increasingly complex decisions with significant ethical implications, raising questions about who is responsible for decisions made by AI and how to ensure that these decisions align with society's ethical and moral values, both in India and the West. Jonathan Haidt has conducted research on moral and ethical decision-making. Today, solving problems like decision-making in autonomous vehicles can draw on the literature of the trolley dilemma in that it illustrates the complexity of (...) ethical decisions faced in emergency situations, as well as the moral implications of the decisions made. However, a series of moral principles must be understood and cited to avoid insufficient mechanistic explanations. Here we present a compilation of ethical positions both in India and the West. (shrink)

When does a model explain? When does it promote understanding? A dominant approach to scientific explanation is the interventionist view (Woodward 2003). According to this view, when X explains Y, intervening on X can produce, prevent or alter Y in some predictable way. In this paper, I argue for two claims. First, I reject a position that many interventionist theorists endorse. This position is that to explain some phenomenon by providing a model is also to understand that phenomenon (Woodward (...) 2003, Ylikoski and Kuorikoski 2010, Kuorikoski and Ylikoski 2015). While endorsing the interventionist view, I argue that explaining and understanding are distinct scientific achievements. Second, I defend a novel theory of scientific understanding. According to this view, when some model M promotes understanding, M makes available a distinctive mental state. This state is of the same psychological kind as when we grasp events in a narrative as bearing on some ultimate conclusion. To conclude, I show that, given this view, mechanistic explanations often provide a powerful source of understanding that many causal-historical models lack. This paper will be of interest to both philosophers of science and epistemologists engaged in the topics of explanation and understanding. (shrink)

This short article is a “conversation” in which an android, Mort, replies to Richard Marc Rubin’s android named Sol in “The Robot Sol Explains Laughter to His Android Brethren” (The Philosophy of Humor Yearbook, 2022). There Sol offers an explanation for how androids can laugh--largely a reaction to frustration and unmet expectations: “my account says that laughter is one of four ways of dealing with frustration, difficulties, and insults. It is a way of getting by. If you need to (...) label my conception of humor, you might call it an adjustment theory, or perhaps an accommodation theory” (Rubin 256). Mort is intrigued but has his doubts. Mort contends that synesthesia likely has something to do with how humans develop humor. Perhaps, he considers, human designers crossed the androids' sensory wires, allowing them to be able to smash up concepts that do not otherwise seem to fit together, reconstruct incongruities in a manner that is not nonsensical, express the aha-eureka moment of a serendipitous discovery, and even feel blue when they fail at a given task that we were expected to complete. (shrink)

This paper integrates type functionalism with the Kairetic account to develop context-specific models for explaining mental states, particularly pain, across different species and systems. By employing context-dependent mapping f_c, we ensure cohesive causal explanations while accommodating multiple realizations of mental states. The framework identifies context subsets C_i and maps them to similarity subspaces S_i, capturing the unique physiological, biochemical, and computational mechanisms underlying pain in different entities such as humans, octopi, and AI systems. This approach highlights the importance of causal (...) relations in defining mental states and preserves their functional roles across diverse contexts. Furthermore, the paper incorporates elements of token functionalism by recognizing species-specific realizations of mental states. By acknowledging the unique representations of mental states within different species and systems, the framework provides a nuanced understanding of how similar functional roles can be fulfilled by diverse physical substrates. This synthesis of type and token functionalism enhances our explanatory power and coherence in addressing the complex nature of mental states. The resulting framework offers a robust tool for analyzing and understanding mental phenomena, with significant implications for cognitive science, philosophy of mind, and artificial intelligence. By maintaining the functional roles of mental states while accommodating their multiple realizations, this approach not only advances theoretical understanding but also opens new avenues for practical applications in cross-species empathy, AI ethics, and the development of context-aware cognitive models. (shrink)

This essay is divided into two parts. In the first part (§2), I introduce the idea of practical meaning by looking at a certain kind of procedural systems — the motor system — that play a central role in computational explanations of motor behavior. I argue that in order to give a satisfactory account of the content of the representations computed by motor systems (motor commands), we need to appeal to a distinctively practical kind of meaning. Defending the explanatory relevance (...) of semantic properties in a computationalist explanation of motor behavior, my argument concludes that practical meanings play a central role in an adequate psychological theory of motor skill. In the second part of this essay (§3), I generalize and clarify the notion of practical meaning, and I defend the intelligibility of practical meanings against an important objection. (shrink)

This paper aims at introducing a French audience to the Intelligent Design debate. It starts by reviewing recent attacks on any possibility of a rational account of theism in light of the contemporary theory of evolution. A section is devoted to outlining the genesis of the "wedge" strategy, to distinguish it from young earth creationism, and to highlight the questioning of evolution as our meta-narrative bearing on overall conceptions of the scientific endeavor. The arguments propounded by Behe are reviewed in (...) detail, as the example of blood coagulation as a purported irreducibly complex mechanism is contrasted to the question of system boundaries. Counter-arguments by Miller are reviewed, and a few bottom-line questions are directed to invite a further consideration of the type of engineering this could testify to. The concept of complex specified information is then presented, and contrasted to traditional and lingering axiomatic problems in the theory of probabilities. The problem of the recognition of an intrinsic pattern is also brought together with some work in recent philosophy of causation and explanation. The two opposing strategies, naturalistic evolutionism and intelligent design, are seen as caught in each other's rhetoric, and some probing is offered in the direction of a theory that could surpass those stalling factors. (shrink)

In this paper I argue that the search for explainable models and interpretable decisions in AI must be reformulated in terms of the broader project of offering a pragmatic and naturalistic account of understanding in AI. Intuitively, the purpose of providing an explanation of a model or a decision is to make it understandable to its stakeholders. But without a previous grasp of what it means to say that an agent understands a model or a decision, the explanatory strategies (...) will lack a well-defined goal. Aside from providing a clearer objective for XAI, focusing on understanding also allows us to relax the factivity condition on explanation, which is impossible to fulfill in many machine learning models, and to focus instead on the pragmatic conditions that determine the best fit between a model and the methods and devices deployed to understand it. After an examination of the different types of understanding discussed in the philosophical and psychological literature, I conclude that interpretative or approximation models not only provide the best way to achieve the objectual understanding of a machine learning model, but are also a necessary condition to achieve post hoc interpretability. This conclusion is partly based on the shortcomings of the purely functionalist approach to post hoc interpretability that seems to be predominant in most recent literature. (shrink)

We argue that general intelligence, as presented in the target article, generates multiple distinct and non-equivalent characterisations. Clarifying this central concept is necessary for assessing Burkart et al.’s proposal that the cultural intelligence hypothesis is the best explanation for the evolution of general intelligence. We assess this claim by considering two characterisations of general intelligence presented in the article.

How many factors, i.e. departures from normality, are necessary to explain a delusion? Maher’s classic one-factor theory argues that the only factor is the patient’s anomalous experience, and a delusion arises as a normal explanation of this experience. The more recent two-factor theory, on the other hand, contends that a second factor is also needed, with reasoning abnormality being a potential candidate, and a delusion arises as an abnormal explanation of the anomalous experience. In the past few years, (...) although there has been an increasing number of scholars offering a variety of arguments in defence of Maher’s one-factor theory, these arguments have not been adequately addressed by two-factor theorists. This paper aims to address this gap by critically examining the arguments on three crucial issues: the intelligibility of delusions, the dissociation between anomalous experiences and delusions, and the empirical evidence of a second factor. I will argue that the Maherian notion of anomalous experience is not sufficient for explaining delusions and the two-factor theory is on the right track in its search for the missing factor in the aetiology of delusions. (shrink)

This essay argues that Huey Newton’s philosophical explanation of US empire fills an epistemological gap in our thinking that provides us with a basis for understanding the emergence and operational application of predictive policing, Big Data, cutting-edge surveillance programs, and semi-autonomous weapons by US military and policing apparati to maintain control over racialized populations historically and in the (still ongoing) Global War on Terror today – a phenomenon that Black Studies scholars and Black philosophers alike have yet to demonstrate (...) the conceptual acumen to explain. Newton’s ability to explain the conceptual foundation of US empire alongside his prediction of the emergence and application of AI and other Big Data technologies towards the repression of dissidents, criminals, terrorists, and other racialized groups abroad gives intercommunalism the explanatory power to lay the basis for a genealogical account of Black repression that also advances our understanding of how the US as a liberal democratic regime engages in enemy target acquisition (the construction of racialized groups as threats internally and externally) to rationalize the brutal repression of these populations by military-police agencies whose operational application of new AI and computing-surveillance technologies and semi-autonomous weapons are conditioned by the intensity of their racialized dehumanization of the enemy. (shrink)

Dear Colleagues, -/- Recently, for an article submitted to the European Journal of Therapeutics, it was reported that the paper may have been written with artificial intelligence support at a rate of more than 50% in the preliminary examination made with Turnitin. However, the authors did not mention this in the article’s material method or explanations section. Fortunately, the article’s out-of-date content and fundamental errors in its methodology allowed us no difficulty making the desk rejection decision. -/- On the other (...) hand, similar situations that we may encounter later caused us to discuss how we would decide when the artificial intelligence support of the articles was written. The general opinion that we have adopted and currently available in the literature is that if artificial intelligence is used while writing an article, how artificial intelligence is used in the methodology should be written in detail. -/- Moreover, we encountered a much more exciting situation during our evaluation. In a few academic studies, we have seen that artificial intelligence is added as a co-author. On July 06, 2023, in the Web of Science, using the advanced search, we found four articles with the author name ChatGPT [1]. We have determined that ChatGPT is the author in one of these articles [2] and the Group Author in three [3-5]. -/- Lee [6] stated that although artificial intelligence tools are much more advanced than search engines, they cannot be an author regarding research ethics because they cannot take responsibility for what they write. Similarly, Goto and Katanoda [7] stated that it is the author’s responsibility to confirm that the texts written by ChatGPT are correct. However, Pourhoseingholi et al. [8] reported that keeping up with technology is inevitable. Additionally, they said that“this action will be more fruitful and practical in extended dimensions when international institutes like ICMJE or COPE come up with the appropriate adjustments and establish robust criteria to scheme the AI authorship”. -/- Most probably, the use of artificial intelligence applications in scientific articles and whether it can be a co-author in these papers will be discussed soon. -/- We encourage interested authors to submit their ideas to our journal as a letter to the editor. -/- Yours sincerely,. (shrink)

The basic task of the essay is to exhibit science as a rational enterprise. I argue that in order to do this we need to change quite fundamentally our whole conception of science. Today it is rather generally taken for granted that a precondition for science to be rational is that in science we do not make substantial assumptions about the world, or about the phenomena we are investigating, which are held permanently immune from empirical appraisal. According to this standard (...) view, science is rational precisely because science does not make a priori metaphysical presuppositions about the world forever preserved from possible empirical refutation. It is of course accepted that an individual scientist, developing a new theory, may well be influenced by his own metaphysical presuppositions. In addition, it is acknowledged that a successful scientific theory, within the context of a particular research program, may be protected for a while from refutation, thus acquiring a kind of temporary metaphysical status, as long as the program continues to be empirically progressive. All such views unite, however, in maintaining that science cannot make permanent metaphysical presuppositions, held permanently immune from objective empirical evaluation. According to this standard view, the rationality of science arises, not from the way in which new theories are discovered, but rather from the way in which already formulated theories are appraised in the light of empirical considerations. And the fundamental problem of the rationality of science--the Humean problem of induction--concerns precisely the crucial issue of the rationality of accepting theories in the light of evidence. In this essay I argue that this widely accepted standard conception of science must be completely rejected if we are to see science as a rational enterprise. In order to assess the rationality of accepting a theory in the light of evidence it is essential to consider the ultimate aims of science. This is because adopting different aims for science will lead us, quite rationally, to accept different theories in the light of evidence. I argue that a basic aim of science is to explain. At the outset science simply presupposes, in a completely a priori fashion, that explanations can be found, that the world is ultimately intelligible or simple. In other words, science simply presupposes in an a priori way the metaphysical thesis that the world is intelligible, and then seeks to convert this presupposed metaphysical theory into a testable scientific theory. Scientific theories are only accepted insofar as they promise to help us realize this fundamental aim. At once a crucial problem arises. If scientific theories are only accepted insofar as they promise to lead us towards articulating a presupposed metaphysical theory, it is clearly essential that we can choose rationally, in an a priori way, between all the very different possible metaphysical theories that can be thought up, all the very different ways in which the universe might ultimately be intelligible. For holding different aims, accepting different metaphysical theories conceived of as blueprints for future scientific theories will, quite rationally, lead us to accept different scientific theories. Thus it is only if we can choose rationally between conflicting metaphysical blueprints for future scientific theories that we will be in a position to appraise rationally the acceptability of our present day scientific theories. We thus face the crucial problem: How can we choose rationally between conflicting possible aims for science, conflicting metaphysical blueprints for future scientific theories? It is only if we can solve this fundamental problem concerning the aims of science that we can be in a position to appraise rationally the acceptability of existing scientific theories. There is a further point here. If we could choose rationally between rival aims, rival metaphysical blueprints for future scientific theories, then we would in effect have a rational method for the discovery of new scientific theories! Thus we reach the result: there is only a rational method for the appraisal of existing scientific theories if there is a rational method of discovery. I shall argue that the aim-oriented theory of scientific inquiry to be advocated here succeeds in exhibiting science as a rational enterprise in that it succeeds in providing a rational procedure for choosing between rival metaphysical blueprints: it thus provides a rational, if fallible, method of discovery, and a rational method for the appraisal of existing scientific theories--thus resolving the Humean problem. In Part I of the essay I argue that the orthodox conception of science fails to exhibit science as a rational enterprise because it fails to solve the Humean problem of induction. The presuppositional view advocated here does however succeed in resolving the Humean problem. In Part II of the essay I spell out the new aim-oriented theory of scientific method that becomes inevitable once we accept the basic presuppositional viewpoint. I argue that this new aim oriented conception of scientific method is essentially a rational method of scientific discovery, and that the theory has important implications for scientific practice. (shrink)

This dissertation argues that cognition is a kind of natural agency. Natural agency is the capacity that certain systems have to act in accordance with their own norms. Natural agents are systems that bias their repertoires in response to affordances in the pursuit of their goals. Cognition is a special mode of this general phenomenon. Cognitive systems are agents that have the additional capacity to actively take their worlds to be certain ways, regardless of whether the world is really that (...) way. In this way, cognitive systems are desituated. Desituatedness is the root of specifically cognitive capacities for representation and abstraction. There are two main reasons why this view needs defending. First, natural agency is typically viewed as incompatible with natural science because it is committed to a teleological mode of explanation. Second, cognition is typically held to be categorically distinct from natural agency. This dissertation argues against both of these views. It argues against the incompatibility of agency and natural science by demonstrating that systems biology, general systems theory, and sciences that deal with complex systems have typically underappreciated conceptual and theoretical resources for grounding agency in the causal structure of the world. These conceptual resources do not, however, reduce agency to systems theory because the normativity inherent in agency demands descriptive resources beyond those of even the most sophisticated systems theory. It argues against the categorical difference between natural agency and cognition by pointing out that separating cognition from a richer web of situated, ecologically embedded relations between the agent and the world generates the frame problem, which is an insuperable obstacle to making cognition that is sufficiently responsive to the complexity of the world. Rooting cognition in natural agency is a more robust empirical bet for theorizing cognition and artificial intelligence. (shrink)

Consciousness may not only be a problem how to know the brain but also a problem how to understand what known. Understanding is always an ontological system created as the explanation of what known by us. And, if all brains, including human brains, may be defined as the mind, consciousness must be part of our understanding of the mind. The author argues that no mind may exist if not be a life or lives, no life may exist if not (...) be a mind or a part of it, and, if it is the mind that needs to be explained, it must finally and fundamentally be explained as a life or a living system. An ontological definition of life is proposed first in this article, which is also applicable to fields such as physics, biology, psychology, and sociology. Based on an analysis of two kinds of lives and their relationship with matter and energy, a living system is modeled as the organization of two system relations, which explains not only consciousness but also wakefulness, emotion, intelligence, language and mind-body relationship. And finally, a semantic theory is reached, which takes lives as the only meaning of all symmetrical and conservative changes, such as the location changes or form changes of consciousness, or the state changes of memories. The concept of life, the model of living systems and the semantic theory together provide a framework for us to understand what and how consciousness is, why there is consciousness, where and when consciousness occurs. (shrink)

It is often thought that traditional recidivism prediction tools used in criminal sentencing, though biased in many ways, can straightforwardly avoid one particularly pernicious type of bias: direct racial discrimination. They can avoid this by excluding race from the list of variables employed to predict recidivism. A similar approach could be taken to the design of newer, machine learning-based (ML) tools for predicting recidivism: information about race could be withheld from the ML tool during its training phase, ensuring that the (...) resulting predictive model does not use race as an explicit predictor. However, if race is correlated with measured recidivism in the training data, the ML tool may ‘learn’ a perfect proxy for race. If such a proxy is found, the exclusion of race would do nothing to weaken the correlation between risk (mis)classifications and race. Is this a problem? We argue that, on some explanations of the wrongness of discrimination, it is. On these explanations, the use of an ML tool that perfectly proxies race would (likely) be more wrong than the use of a traditional tool that imperfectly proxies race. Indeed, on some views, use of a perfect proxy for race is plausibly as wrong as explicit racial profiling. We end by drawing out four implications of our arguments. (shrink)

Abstract: This paper argues that a central explanatory role for the concept of Umwelt in theoretical biology is to be found in developmental biology, in particular in the effort to understand development as a goal-directed and adaptive process that is controlled by the organism itself. I will reach this conclusion in two (interrelated) ways. The first is purely theoretical and relates to the current scenario in the philosophy of biology. Challenging neo-Darwinism requires a new understanding of the various components involved (...) in natural selection processes. An important prerequisite for the explanation is the ability to understand development in a teleological way. Here, the concept of Umwelt plays a crucial role: if organisms are responsible for generating adaptive variation in specific environments, we need a theory that explains the context-dependent nature of adaptively oriented processes. The Umwelt is thus a central element in determining the goal that an adaptive process pursues. The second path in my analysis also has a historical dimension. I will present Karl Ernst von Baer’s reflections on teleological development and his influence on von Uexküll's thinking. I will present various ideas developed by von Baer, such as the distinction between Ziel and Zweck and the use of musical metaphors, which can help to understand development teleologically and give von Uexküll’s theory a central place in this framework. (shrink)

The notion of context arises in assorted areas of artificial intelligence (AI), including knowledge representation, natural language processing, intelligent information retrieval, etc. Although the term ‘context’ is frequently employed in descriptions, explanations, and analyses of computer programs in these areas, its meaning is frequently left to the reader’s understanding. -/- My aim in this paper is to offer a swift review of context in AI. I will first identify the role of context in various fields of AI. I will then (...) consider the enterprise of formalising context. The focus of my discussion will be McCarthy's (1993) proposal – a landmark study that inspired almost every logical theory of context since then – and its close variants, studied by his students. -/- (The author is grateful to Prof. Carlo Penco for his invaluable assistance with this work.). (shrink)

In this paper, I shall consider the challenge that Quine posed in 1947 to the advocates of quantified modal logic to provide an explanation, or interpretation, of modal notions that is intuitively clear, allows “quantifying in”, and does not presuppose, mysterious, intensional entities. The modal concepts that Quine and his contemporaries, e.g. Carnap and Ruth Barcan Marcus, were primarily concerned with in the 1940’s were the notions of (broadly) logical, or analytical, necessity and possibility, rather than the metaphysical modalities (...) that have since become popular, largely due to the influence of Kripke. In the 1950’s modal logicians responded to Quine’s challenge by providing quantified modal logic with model-theoretic semantics of various types. In doing so they also, explicitly or implicitly addressed Quine’s interpretation problem. Here I shall consider the approaches developed by Carnap in the late 1940’s, and by Kanger, Hintikka, Montague, and Kripke in the 1950’s, and discuss to what extent these approaches were successful in meeting Quine’s doubts about the intelligibility of quantified modal logic. (shrink)

In this paper, I examine Wittgenstein’s conception of reason and rationality through the lens of his conception of reasons. Central in this context, I argue, is the image of the chain, which informs not only his methodology in the form of the chain-method, but also his conception of reasons as linking up immediately, like the links of a chain. I first provide a general sketch of what reasons are on Wittgenstein’s view, arguing that giving reasons consists in making thought and (...) action intelligible by delineating reasoning routes; that something is a reason not in virtue of some intrinsic property, but in virtue of its role; and that citing something as a reason characterises it in terms of the rational relations it stands in according to context-dependent norms. I then argue that on Wittgenstein’s view, we misconceive chains of reasons if we think of them on the model of chains of causes. Chains of reasons are necessarily finite, because they are anchored in and held in place by our reason-giving practices, and it is in virtue of their finitude that chains of reasons can guide, justify and explain. I argue that this liberates us from the expectation that one should be able to give reasons for everything, but that it limits the reach of reasons by tying them to particular reasoning-practices that they cannot themselves justify. I end by comparing and reconciling Wittgenstein’s dichotomy between chains of reasons and chains of causes with seemingly competing construals of the dichotomy, and I clarify its relation to the dichotomy between explanation and justification. (shrink)

The main thesis of Michael Della Rocca’s outstanding Spinoza book (Della Rocca 2008a) is that at the very center of Spinoza’s philosophy stands the Principle of Sufficient Reason (PSR): the stipulation that everything must be explainable or, in other words, the rejection of any brute facts. Della Rocca rightly ascribes to Spinoza a strong version of the PSR. It is not only that the actual existence and features of all things must be explicable, but even the inexistence – as well (...) as the absence of any feature of any thing – demands an explanation. Della Rocca does not stop here, however. He feeds his PSR monster with some more powerful steroids and suggests that Spinoza advocates what he terms “the twofold use of the PSR.” It is not only that everything must be explained and made intelligible, but it must ultimately be explained in terms of explainability or intelligibility itself. This twofold use of the PSR is the key to the entire book. Della Roca’s strategy throughout the book is to argue that any key feature of Spinoza’s system – be it causality, inherence, essence, consciousness, existence, rejection of teleology, goodness or political right – must be explained, and ultimately it must be explained in terms of intelligibility. “Spinoza single-mindedly digs and digs until we find that the phenomenon in question is nothing but some form of intelligibility itself, of explicability itself” (Della Rocca 2008a: 2). Della Rocca’s book came out together with a cluster of articles in which he develops in detail his new reading of Spinoza. In one of these articles, he warns the reader: “Don’t let me start” (Dell Rocca 2010: 1). The train that is about to embark leads to very bizarre terrain, and thus one should think twice before embarking on the “PSR Express.” In this paper I argue that the train was hijacked. This was a perfect crime: without anyone noticing it, the engine driver diverted the train to a new route, and as with other perfect crimes, it is none but the criminal himself who is capable of, and indeed will, bring about his own demise. As I will later argue, Della Rocca’s “PSR-on-steroids” will eventually cripple reason itself. But let us not run too fast, and start at the very beginning. I happily – or at least, so I think - board the “PSR Express.” I believe Spinoza is strongly committed to the PSR and makes very significant use of this principle, but, unlike Della Rocca, I do not think the PSR is the key to all mysteries Spinozist, nor do I believe Spinoza was committed to the reductionist program of explaining all things through intelligibility (i.e., the second use of the PSR). (shrink)

The use of counterfactuals for considerations of algorithmic fairness and explainability is gaining prominence within the machine learning community and industry. This paper argues for more caution with the use of counterfactuals when the facts to be considered are social categories such as race or gender. We review a broad body of papers from philosophy and social sciences on social ontology and the semantics of counterfactuals, and we conclude that the counterfactual approach in machine learning fairness and social explainability can (...) require an incoherent theory of what social categories are. Our findings suggest that most often the social categories may not admit counterfactual manipulation, and hence may not appropriately satisfy the demands for evaluating the truth or falsity of counterfactuals. This is important because the wide- spread use of counterfactuals in machine learning can lead to misleading results when applied in high-stakes domains. Accordingly, we argue that even though counterfactuals play an essential part in some causal inferences, their use for questions of algorithmic fairness and social explanations can create more problems than they resolve. Our positive result is a set of tenets about using counterfactuals for fairness and explanations in machine learning. (shrink)